Tutorials#
Each tutorial solves one task with a script that runs from top to bottom on a CPU. The documentation build runs every script, so the output on each page comes from the code above it. Each page has links to download the script and a Jupyter notebook, and ends with a “Further reading” list of the guides and API pages that go deeper. If you are new to tensordict, read Getting started first.
The sidebar lists the tutorials in four groups, in reading order:
Basics: keys, indexing, shapes, row-by-row filling, and lazy stacks of streamed data.
Storage: devices and memory-mapped tensors, the zarr backend, and a benchmark of the serialization formats.
tensordict.nn: modules that read and write tensordicts, functional calls and ensembles, and export with
torch.export.Data loading: datasets in a
TensorDictor aTensorClass, and batched loading of an image dataset.
